linux-stable.git/arch/arm, branch v3.2.92

mm: larger stack guard gap, between vmas

2017-07-02T16:12:47+00:00

commit 1be7107fbe18eed3e319a6c3e83c78254b693acb upstream.

Stack guard page is a useful feature to reduce a risk of stack smashing
into a different mapping. We have been using a single page gap which
is sufficient to prevent having stack adjacent to a different mapping.
But this seems to be insufficient in the light of the stack usage in
userspace. E.g. glibc uses as large as 64kB alloca() in many commonly
used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX]
which is 256kB or stack strings with MAX_ARG_STRLEN.

This will become especially dangerous for suid binaries and the default
no limit for the stack size limit because those applications can be
tricked to consume a large portion of the stack and a single glibc call
could jump over the guard page. These attacks are not theoretical,
unfortunatelly.

Make those attacks less probable by increasing the stack guard gap
to 1MB (on systems with 4k pages; but make it depend on the page size
because systems with larger base pages might cap stack allocations in
the PAGE_SIZE units) which should cover larger alloca() and VLA stack
allocations. It is obviously not a full fix because the problem is
somehow inherent, but it should reduce attack space a lot.

One could argue that the gap size should be configurable from userspace,
but that can be done later when somebody finds that the new 1MB is wrong
for some special case applications.  For now, add a kernel command line
option (stack_guard_gap) to specify the stack gap size (in page units).

Implementation wise, first delete all the old code for stack guard page:
because although we could get away with accounting one extra page in a
stack vma, accounting a larger gap can break userspace - case in point,
a program run with "ulimit -S -v 20000" failed when the 1MB gap was
counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK
and strict non-overcommit mode.

Instead of keeping gap inside the stack vma, maintain the stack guard
gap as a gap between vmas: using vm_start_gap() in place of vm_start
(or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few
places which need to respect the gap - mainly arch_get_unmapped_area(),
and and the vma tree's subtree_gap support for that.

Original-patch-by: Oleg Nesterov 
Original-patch-by: Michal Hocko 
Signed-off-by: Hugh Dickins 
Acked-by: Michal Hocko 
Tested-by: Helge Deller  # parisc
Signed-off-by: Linus Torvalds 
[Hugh Dickins: Backported to 3.2]
[bwh: Fix more instances of vma->vm_start in sparc64 impl. of
 arch_get_unmapped_area_topdown() and generic impl. of
 hugetlb_get_unmapped_area()]
Signed-off-by: Ben Hutchings

ARM: 8643/3: arm/ptrace: Preserve previous registers for short regset write

2017-03-16T02:18:47+00:00

commit 228dbbfb5d77f8e047b2a1d78da14b7158433027 upstream.

Ensure that if userspace supplies insufficient data to
PTRACE_SETREGSET to fill all the registers, the thread's old
registers are preserved.

Fixes: 5be6f62b0059 ("ARM: 6883/1: ptrace: Migrate to regsets framework")
Signed-off-by: Dave Martin 
Acked-by: Russell King 
Signed-off-by: Russell King 
Signed-off-by: Ben Hutchings

ARM: 8634/1: hw_breakpoint: blacklist Scorpion CPUs

2017-03-16T02:18:42+00:00

commit ddc37832a1349f474c4532de381498020ed71d31 upstream.

On APQ8060, the kernel crashes in arch_hw_breakpoint_init, taking an
undefined instruction trap within write_wb_reg. This is because Scorpion
CPUs erroneously appear to set DBGPRSR.SPD when WFI is issued, even if
the core is not powered down. When DBGPRSR.SPD is set, breakpoint and
watchpoint registers are treated as undefined.

It's possible to trigger similar crashes later on from userspace, by
requesting the kernel to install a breakpoint or watchpoint, as we can
go idle at any point between the reset of the debug registers and their
later use. This has always been the case.

Given that this has always been broken, no-one has complained until now,
and there is no clear workaround, disable hardware breakpoints and
watchpoints on Scorpion to avoid these issues.

Signed-off-by: Mark Rutland 
Reported-by: Linus Walleij 
Reviewed-by: Stephen Boyd 
Acked-by: Will Deacon 
Cc: Russell King 
Signed-off-by: Russell King 
[bwh: Backported to 3.2:
 - Open-code read_cpuid_part()
 - Adjust context]
Signed-off-by: Ben Hutchings

ARM: dma-mapping: don't allow DMA mappings to be marked executable

2017-02-23T03:51:03+00:00

commit 0ea1ec713f04bdfac343c9702b21cd3a7c711826 upstream.

DMA mapping permissions were being derived from pgprot_kernel directly
without using PAGE_KERNEL.  This causes them to be marked with executable
permission, which is not what we want.  Fix this.

Signed-off-by: Russell King 
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings

ARM: sa1111: fix pcmcia suspend/resume

2016-11-20T01:01:38+00:00

commit 06dfe5cc0cc684e735cb0232fdb756d30780b05d upstream.

SA1111 PCMCIA was broken when PCMCIA switched to using dev_pm_ops for
the PCMCIA socket class.  PCMCIA used to handle suspend/resume via the
socket hosting device, which happened at normal device suspend/resume
time.

However, the referenced commit changed this: much of the resume now
happens much earlier, in the noirq resume handler of dev_pm_ops.

However, on SA1111, the PCMCIA device is not accessible as the SA1111
has not been resumed at _noirq time.  It's slightly worse than that,
because the SA1111 has already been put to sleep at _noirq time, so
suspend doesn't work properly.

Fix this by converting the core SA1111 code to use dev_pm_ops as well,
and performing its own suspend/resume at noirq time.

This fixes these errors in the kernel log:

pcmcia_socket pcmcia_socket0: time out after reset
pcmcia_socket pcmcia_socket1: time out after reset

and the resulting lack of PCMCIA cards after a S2RAM cycle.

Fixes: d7646f7632549 ("pcmcia: use dev_pm_ops for class pcmcia_socket_class")
Signed-off-by: Russell King 
Signed-off-by: Ben Hutchings

arm: oabi compat: add missing access checks

2016-11-20T01:01:32+00:00

commit 7de249964f5578e67b99699c5f0b405738d820a2 upstream.

Add access checks to sys_oabi_epoll_wait() and sys_oabi_semtimedop().
This fixes CVE-2016-3857, a local privilege escalation under
CONFIG_OABI_COMPAT.

Reported-by: Chiachih Wu 
Reviewed-by: Kees Cook 
Reviewed-by: Nicolas Pitre 
Signed-off-by: Dave Weinstein 
Signed-off-by: Linus Torvalds 
Signed-off-by: Ben Hutchings

ARM: OMAP3: hwmod data: Add sysc information for DSI

2016-11-20T01:01:28+00:00

commit b46211d6dcfb81a8af66b8684a42d629183670d4 upstream.

Add missing sysconfig/sysstatus information
to OMAP3 hwmod. The information has been
checked against OMAP34xx and OMAP36xx TRM.

Without this change DSI block is not reset
during boot, which is required for working
Nokia N950 display.

Signed-off-by: Sebastian Reichel 
Signed-off-by: Tony Lindgren 
Signed-off-by: Ben Hutchings

ARM: fix PTRACE_SETVFPREGS on SMP systems

2016-08-22T21:37:13+00:00

commit e2dfb4b880146bfd4b6aa8e138c0205407cebbaf upstream.

PTRACE_SETVFPREGS fails to properly mark the VFP register set to be
reloaded, because it undoes one of the effects of vfp_flush_hwstate().

Specifically vfp_flush_hwstate() sets thread->vfpstate.hard.cpu to
an invalid CPU number, but vfp_set() overwrites this with the original
CPU number, thereby rendering the hardware state as apparently "valid",
even though the software state is more recent.

Fix this by reverting the previous change.

Fixes: 8130b9d7b9d8 ("ARM: 7308/1: vfp: flush thread hwstate before copying ptrace registers")
Acked-by: Will Deacon 
Tested-by: Simon Marchi 
Signed-off-by: Russell King 
Signed-off-by: Ben Hutchings

ARM: 8519/1: ICST: try other dividends than 1

2016-02-27T14:28:48+00:00

commit e972c37459c813190461dabfeaac228e00aae259 upstream.

Since the dawn of time the ICST code has only supported divide
by one or hang in an eternal loop. Luckily we were always dividing
by one because the reference frequency for the systems using
the ICSTs is 24MHz and the [min,max] values for the PLL input
if [10,320] MHz for ICST307 and [6,200] for ICST525, so the loop
will always terminate immediately without assigning any divisor
for the reference frequency.

But for the code to make sense, let's insert the missing i++

Reported-by: David Binderman 
Signed-off-by: Linus Walleij 
Signed-off-by: Russell King 
Signed-off-by: Ben Hutchings

ARM: 8517/1: ICST: avoid arithmetic overflow in icst_hz()

2016-02-27T14:28:48+00:00

commit 5070fb14a0154f075c8b418e5bc58a620ae85a45 upstream.

When trying to set the ICST 307 clock to 25174000 Hz I ran into
this arithmetic error: the icst_hz_to_vco() correctly figure out
DIVIDE=2, RDW=100 and VDW=99 yielding a frequency of
25174000 Hz out of the VCO. (I replicated the icst_hz() function
in a spreadsheet to verify this.)

However, when I called icst_hz() on these VCO settings it would
instead return 4122709 Hz. This causes an error in the common
clock driver for ICST as the common clock framework will call
.round_rate() on the clock which will utilize icst_hz_to_vco()
followed by icst_hz() suggesting the erroneous frequency, and
then the clock gets set to this.

The error did not manifest in the old clock framework since
this high frequency was only used by the CLCD, which calls
clk_set_rate() without first calling clk_round_rate() and since
the old clock framework would not call clk_round_rate() before
setting the frequency, the correct values propagated into
the VCO.

After some experimenting I figured out that it was due to a simple
arithmetic overflow: the divisor for 24Mhz reference frequency
as reference becomes 24000000*2*(99+8)=0x132212400 and the "1"
in bit 32 overflows and is lost.

But introducing an explicit 64-by-32 bit do_div() and casting
the divisor into (u64) we get the right frequency back, and the
right frequency gets set.

Tested on the ARM Versatile.

Cc: linux-clk@vger.kernel.org
Cc: Pawel Moll 
Signed-off-by: Linus Walleij 
Signed-off-by: Russell King 
Signed-off-by: Ben Hutchings